• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Question Intel Mont thread

Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
LightningZ71 said:
Leaving fantasy land, we do know that a few things are in play:
Intel is going to begin to retire processors from Intel 7 (alder/raptor) so there will be excess capacity there. Intel does not currently have an Intel 3 based IMC design/implementation. They aren't going to invest the money in making one for an N-series processor.

Skymont, as implemented, only exists on expensive TSMC logic. They won't pay for an N series chip on a leading TSMC node. We won't see a Skymont N series product any time soon, or possibly ever.

What we will see is an N series refresh of the current Alder Lake based product. You might get higher clocks, but that's about it. I would not be shocked if they used the 4 X 4 Crestmont cluster from Raptor Lake in a 16 e-core embedded product.
Click to expand...
You are right, I totally forgot that detail!
Unless TSMC offers a discount for that in order to not to let Intel to use their foundries.
 
Atom Skymont vs Zen5c

8c/8T Atom Skymont 3.6GHz - 4.6GHz (Intel Core Ultra 5 245K (14C/14T, 125-159W)
• CPU Multi Thread : 6 005,09 (+13,46%)
• CPU Single Thread : 753,2 (+45,13%)


8c/16T Zen5c 2.0GHz - 3.3GHz (AMD Ryzen AI 9 HX 370 12C/24T, 15-54W)
• CPU Multi Thread : 5 294 (88,15%)
• CPU Single Thread : 519 (68,91%)


source: https://diit.cz/clanek/recenze-intel-core-ultra-5-245k-aneb-levnejsi-arrow-lake-v-akci
source: https://diit.cz/clanek/preview-asus-zenbook-s16-s-amd-ryzen-ai-9-hx-370-je-v-redakci
 
Damn... Skymont is hella good!

And makes me think if Intel will release the Alder Lake succesor in a future.
 
moinmoin said:
@Kosusko
3.6GHz - 4.6GHz @ 125-159W vs. 2.0GHz - 3.3GHz @ 15-54W?
That's quite the silly comparison.
Click to expand...
125W and 159W are just the TDP and PL2 of a complete 245K CPU. Obviously, it will be much lower power if the CPU is running with the P cores disabled.

You really need the actual power consumption of both during that workload to draw a conclusion.

Regardless, SKT winning in MT while having just 8C/8T and being such a compact core is pretty solid.
 
Never mind, mobile Arrow Lake H/HX with atomic Skymont E-cores are here in Q1/2025... and then we'll get back to it...
 
I'd like to see a 4 cluster Skymont (16 core) cpu. What's the point of Lion Cove in mobile? Ultra mobile frequency tops out below 5GHz in real world usage and Skymont can do that with much better area efficiency compared to Lion Cove. Even at 3.5GHz such a part would do 20,000 CB R23 MT.
 
Hulk said:
What's the point of Lion Cove in mobile?
Click to expand...
ST perf.
Hulk said:
Ultra mobile frequency tops out below 5GHz in real world usage
Click to expand...
Is this real world usage with us in the room.
Hulk said:
Even at 3.5GHz such a part would do 20,000 CB R23 MT.
Click to expand...
Turns out mobile chips are for anything but cinememe.
They tend to do stuff like web browsing and all. Stuff that needs a big core with maximum 1t perf available.
 
adroc_thurston said:
ST perf.

Is this real world usage with us in the room.

Turns out mobile chips are for anything but cinememe.
They tend to do stuff like web browsing and all. Stuff that needs a big core with maximum 1t perf available.
Click to expand...
Think about it some more. I'll give you a hint, when you remove LNC clock speed advantage (mobile) LNC has very little IPC over Skymont, but is over 3 times larger.
 
Hulk said:
LNC has very little IPC over Skymont, but is over 3 times larger.
Click to expand...
Were it not for their internal politics, they could've gone with Skymont Big and Skymont Little, the former optimized for hitting 5.5 GHz and the latter for conserving power, using the relevant transistor libraries.

This seems to be Lisa's strength. She kicks out people who play politics (like Raja Koduri).
 
Hulk said:
Think about it some more. I'll give you a hint, when you remove LNC clock speed advantage (mobile) LNC has very little IPC over Skymont, but is over 3 times larger.
Click to expand...
igor_kavinski said:
Were it not for their internal politics, they could've gone with Skymont Big and Skymont Little, the former optimized for hitting 5.5 GHz and the latter for conserving power, using the relevant transistor libraries.
Click to expand...

Yeah, if the next small core really only needs to do just one of the 2 things:

1. Provide a 10% bigger IPC uplift than the Lion Cove successor (say 20% vs 10%)
2. Clock to ~5.5 Ghz in a "bigger core" format

If it does any of that (and Intel should already know the simulated performance of both), there is absolutely no techical reason to keep the big core around anywhere but some exotic Server (and perhaps desktop or HEDT) SKUs that need AVX-512, HT, etc ...
 
Hulk said:
Think about it some more. I'll give you a hint, when you remove LNC clock speed advantage (mobile) LNC has very little IPC over Skymont, but is over 3 times larger.
Click to expand...
No core should be 3x the size just to hit 25% more clocks. I'm sure if you gave the E core team just 2x the transistor budget, not even 3x, they'd be able to beat LNC handily in all metrics.
 
Saylick said:
they'd be able to beat LNC handily in all metrics.
Click to expand...
They'd need to increase IPC and clock rate without exploding area. In any case it would no longer be Skymont but something else. And it wouldn't have been ready in time to launch in 2024.
 
Saylick said:
No core should be 3x the size just to hit 25% more clocks. I'm sure if you gave the E core team just 2x the transistor budget, not even 3x, they'd be able to beat LNC handily in all metrics.
Click to expand...
Skymont is a good effort, but it needs to do better. It needs to be in ballpark of ARM.

So Arctic Wolf being 1.3x perf/clock with 1.3x power and area will do that. Then the successor to that needs to do another 1.3x on top of that, and once that is achieved, expand it to get it higher.

Why not now? Because the limits placed on the design forces them to be efficient. Transistor efficient, area efficient, power efficient. Maximize that efficiency first.
 
Kosusko said:

Skymont in Desktop Form: Atom Unleashed​

Skymont is Intel's newest E-Core architecture​

source: https://chipsandcheese.com/p/skymont-in-desktop-form-atom-unleashed​

Click to expand...
Summary from your article link:

The overall performance picture though, is that Skymont is still not at the same level as a high performance core. It’s behind in SPEC CPU2017’s integer suite, which represents a favorable set of workloads for a wide core with relatively weak vector execution. In vector-heavy workloads, it’s still hit or miss against Zen 2. Thus Skymont lacks the performance to replace Lion Cove or take on AMD’s best.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top